Correction for range restriction: Lessons from 20 research scenarios

Data are often available only for recruits, a range-restricted sample. This creates the potential for mistaken inferences and poor decisions. This is because inferences and decisions are about the population, not the sample. Despite these problems, researchers must try to determine statistical values as if the sample was not range-restricted. Although range restriction correction methods have been available for over a century, often they are not applied or are applied incorrectly. Technical psychometric discussions of range restriction have not improved researcher practice. As an alternative, realistic scenarios are presented to illustrate and explain the consequences of (1) failing to correct correlations, (2) using the wrong correction formula, (3) correcting when information about previous selection variables is unavailable, (4) using an inappropriate unrestricted sample, (5) incorrectly computing the confidence interval for corrected correlations, and (6) interpretation of results. Although there are situations under which correction has little effect, correction still provides better estimates of relations among variables. It also improves theoretical understanding and interpretation of real-world results.